OpenAI Unveils Jalapeno ASIC and Gluon Language

The custom AI chip features HBM4 memory and outperforms NVIDIA GB200 systems in efficiency and latency on the InferenceX benchmark.

OpenAI has unveiled the Jalapeno ASIC, a custom chip paired with its Gluon kernel programming language. Built on TSMC’s N3P node with an N3E I/O chiplet, the hardware features HBM4 memory, MXFP4 support, and a single-token prediction architecture for AI tasks.

The chip delivers 13.4 PFLOPs of MXFP4 performance and 15.4 TB/s of memory bandwidth. Utilizing weight-stationary systolic arrays and 64-bit Out-of-Order scalar cores, it maintains a 700W TDP while typically consuming 550W during active inference workloads.

Jalapeno outperforms NVIDIA GB200 and GB300 systems, which use HBM3E. On the InferenceX benchmark, it achieved 1.5x to 1.9x better efficiency per watt and 1.7x to 3.6x lower latency, marking a major shift in high-performance AI compute and power efficiency.


More articles

NVIDIA GeForce NOW Adds DLSS 4.5 and Steam Deck App

Asus Opens ROG Ally X20 Pre-Orders for $1,299.99

ASUS Unveils TUF Gaming RX 9070 XT Onimusha Edition

ASUS ROG Unveils 720Hz OLED Gaming Monitors

Apple Unveils M5 Ultra Chip and New Mac Studio

Arm Ships 1.5 Billion Neoverse Cores to Datacenters

Intel Unveils 256-Core Diamond Rapids Processor

Nine Indicted for Smuggling Nvidia GPUs to China

Intel Developing Wildcat Lake Mobile Processors

Leaked Image Shows Apple M5 Private Cloud Servers

Intel Unveils New Agentic AI Architectures

AMD and Intel Desktop CPU Shipments Drop in Q2

MINISFORUM Launches M2 PRO-358H AI PC in Japan

IBM Unveils Next-Gen Z Chip with Dual Arm ISA

Modder Attaches CPU Coolers to Nvidia GTX GPUs

Micron Unveils HBM4 With 2800 GB/s Bandwidth

AMD Radeon RX 10800 XT Leak Suggests 48GB GDDR7

Singularity Computers Debuts Spectre Limited Case

Chinese Robots Break Usain Bolt's 100m Record

AMD Launches Ryzen 7 7700X3D Gaming CPU for $329